K-means clustering output matrix C (represent centroid locations)

Hello,

I am using kmeans clustering function to cluster a principal components matrix (pcs) with dimension 13149x10, where each column of the pcs matrix represent a principal component with decreasing order of explained variance. The number of clusters that I want to use is k = 8. I want to plot the location of the 8 centroids resulting from the kmeans algorithm, but the output matrix C (which gives the location of the centroids) has a size of 8x10 (k-by-p). Why there are 10 columns/ooordinates in C to define the position of each centroid? It shouldn't it be 2 columns?

The command that I am using is this:

[idx,C,sumd,D] = kmeans(pcs(:,1:10),k,'Replicates',10,'Start','cluster');

Appreciate your help.

Thank you!

Answer 1

[the cyclist]

You have 8 centroids, each of which is a point in a 10-dimensional space. Each row gives the 10 coordinates needed to locate that centroid in 10 dimensions.

I'm not sure why you expected 2 columns.

[Rui Silva]

Thanks for your reply,

I was actually making some confusion with the examples in the documentation which are in two-dimensional space. Now it makes sense.

Kind regards,

Rui Silva